Modeling of Hybrid Henry Gas Solubility Optimization Algorithm with Deep Learning-Based LED Driver System

Abstract

Light emitting diodes (LEDs) have become an effective lighting solution because of the characteristics of energy efficiency, flexible controllability and extended lifetime. They find use in numerous lighting systems for residents, industries, enterprises and street lighting applications. The efficiency and trustworthiness of the LED systems considerably based on the thermal mechanical loading improved several degradation schemes and respective interfaces. The complication of the LED systems limits the theoretic interpretation of the core reasons for the luminous variation or the formation of the direct correlation among the thermal aging loading and the luminous output. Therefore, this paper designs a new hybrid Henry gas solubility optimization with deep learning (HHGSO-DL) algorithm for LED driver system design. The presented HHGSO-DL technique mainly concentrates on the derivation of empirical relationships among the design parameters, thermal aging loading and luminous outcomes of the LED product. In the presented HHGSO-DL technique, bidirectional long short-term memory (BiLSTM) algorithm is executed for examining the empirical relationship and its hyperparameters can be tuned by the HHGSO algorithm. In this work, the HHGSO algorithm is derived by the integration of traditional HGSO algorithm with oppositional-based learning (OBL) concept. The performance of the HHGSO-DL technique can be investigated on LED chip packaging and LED luminaire with thermal aging loading. The extensive results demonstrate the promising performance of the HHGSO-DL technique over other state-of-the-art approaches.